Difficulty of Complicated ModelsIn this page, "statistical models" includes the fields of machine learning and system identification.
Random forest is an approach to make many models. And the approach, "serching the best statistical model", is not difficult today.
Because of that some people say, "It is easy to get models" and "Data scientists will not be needed because the computer find the best model automatically."
But I do not think that there is a statistical model to find the law (formulation) of the nature.
If there are not any hypothesises of formulation, it is very difficult problem.
If there are some hypothesises of the formulation, statistical model is powerful to test the hypothesis.
Because the most data has errors, the data does not fit the model. Even if there are errors, we confirm the model by statistics.
The statistical model is one of the tools to join the data and the model. Especially, the analysis of the propagation of errors is useful.
Another is dimension analysis.
By the analysis, we can get the idea of the suitable model.
NEXT Difficulty of Complicated Models
